History of STV in Canada
Most Canadians don’t know that STV has a distinguished history of use in Canada over the past century. In fact, it’s the only form of PR that has ever been used here and in our neighbour to the south.
At the time it was introduced, it was often referred to simply as “proportional representation” (PR), as most of the other PR systems now used elsewhere in the world hadn’t been invented yet.
STV was first used to elect city councillors in Calgary in 1916 and remained in use until 1974.1 In 1917, the BC legislature allowed cities to adopt STV and several cities there used the system in the late 1910s and through the 1920s, including Vancouver, South Vancouver, Victoria, Nelson, Port Coquitlam, Mission City, New Westminster and West Vancouver. Lethbridge, Edmonton, Regina, Moose Jaw, Saskatoon, North Battleford, Winnipeg, Transcona, St. James and St. Vital also used STV at some time, with the Manitoba cities continuing to use it until 1971. Ottawa city council tried to introduce STV in 1916, but were blocked from doing so by the Ontario legislature.2
At the provincial level, STV was used to elect legislators in Winnipeg from 1920-55 and in Edmonton and Calgary from the 1920s to the 1950s.
In the USA, STV was used in over twenty cities starting about 1915, including such notable cities as New York, Cleveland, Cincinnati, Sacramento and Boulder. Cambridge and Minneapolis continue to use it today.
For the most part, STV was abandoned in most of these places (typically over strong objections of the opposition) because it was very effective in ensuring that minority voices won political representation, in proportion to their voter support. These battles were particularly nasty in the USA where the intent was to marginalize African-Americans and socialists.
More recently, STV has been recommended four times for use in Canada:
- In 1978, the Manitoba Law Reform Commission’s Working Group recommended adopting STV in urban areas and the Alternative Vote in rural areas for provincial elections.
- In 2004, the BC Citizens’ Assembly chose STV for BC provincial elections. BC voters approved the Citizens’ Assembly’s recommendation with 58 percent voting in favour in the 2005 referendum, but failed to reach the super-majority threshold of 60 percent established by the government.
- In 2006, then-Prime Minister Harper tabled Senate reform legislation that proposed to use STV to elect senators.
- On June 9, 2016, Ontario’s provincial government passed legislation that allows Ontario municipalities to elect councillors using ranked ballots; in municipalities with multi-member wards, this means STV.
STV is currently used in Ireland, Malta, two Australian states and territories (Tasmania and ACT), the Australian Senate (although the Australian Senate uses a form of STV that is akin to a party list system), New Zealand local elections and health boards, and Scottish local elections. It is the preferred proportional system of the UK Electoral Reform Society.
How Does STV Work? The Basic Idea
With STV, voters elect a team of local MPs using a ranked ballot.
With STV, groups of like-minded voters elect their preferred candidates. Since neighbours might not agree politically, each geographic area would elect more than one MP. For example, five adjacent ridings that each currently elect one MP could be joined and the voters there would elect their five favourite candidates. Any candidate who earns a defined share of voter support or quota will be sure of earning a seat – 16 percent or more in a five-seat district, although fewer is sometimes enough.
The number of MPs elected per district would vary to suit the geography of the area. This means that some districts would deliver more proportionality than others, but the overall results across the country would be very proportional.
Each voter has one vote using a ranked ballot. The ranked ballot ensures that a vote for a candidate who is eliminated or who already has met the quota can be redistributed to the voter’s next preferred candidate. Voters can rank as many or as few candidates as they wish, and are allowed to vote across parties if they wish. The number of names on a ballot is variable, but is manageable for STV ridings of modest size, keeping in mind that voters are likely to focus on the limited number of candidates from their preferred parties. The sample ballot below illustrates what we might expect in a four member district:
How Elections Canada Counts the Ballots
The simplest way to count STV ballots is the multi-member instant runoff method. Assuming that not all of the positions are filled in the first round, to make second preferences count under this method, the last-place candidates are eliminated and their votes are redistributed to second preferences. This process is repeated until there are only as many candidates left as seats. These are declared elected as MPs. This approach is the same as the AV system, except that it is applied in a multi-member context.
While this simple run-off counting method is possible, most places using STV complement this with measures to redistribute any “surplus” votes, as well, to ensure that more votes count equally. This method is used in Ireland and was the one recommended by the BC Citizens’ Assembly:
- The core principle is that each elected candidate should have the support of roughly the same number of voters. This fair share of votes is called a ‘quota’ and represents the total number of votes cast divided by the number of seats plus one. For example, if there are 1000 voters and four seats, the quota will be 1000/(4+1)+1 = 201 votes.
- Any candidate who reaches the quota wins a seat. In each round of counting, if one or more candidates have more votes than the quota, the extra votes are transferred to the next choice on each ballot.
- The runoff rule is then applied by eliminating the candidate with the fewest votes and transferring their votes to the next choices on each ballot.
These steps are repeated until all of the seats have been filled.
What Could STV Look Like in Canada?
Given Canada’s significant practical experience with STV, what would STV look like in Canada?
In its most basic form, STV would group together neighbouring ridings and voters would elect a team of local MPs to collectively represent the voters. The number of MPs in a district could vary from as few as two, to as many as ten or more (for example, Winnipeg elected ten MPs to the provincial legislature from a single citywide district for the first 25 years that STV was used there).
It would be up to an Electoral Boundaries Commission to make the final decision in each region about how many seats are put into a particular district, taking into account the input they would receive during their normal public consultation process.
The BC Citizens’ Assembly recommended that the most rural parts of the province should have the smallest districts (as few as two or three seats), and the most urban areas should have the largest (as many as six or seven).
Here, we show examples of how STV districts might look in two different parts of the country – Vancouver Island and downtown Toronto.
Vancouver Island currently elects seven MPs, as shown in Figure 1, and represents three distinct regions – a sparsely populated north, a string of smaller cities and towns mainly along the east coast mid-island, and the metro Victoria region in the south.
The population dot map in Figure 2 shows how the population of the Island is distributed. If we had conducted the 2015 election under STV, the Electoral Boundaries Commission (EBC) might have opted to have two electoral districts – one for the northern portion of Vancouver Island and one for the southern portion, so the EBC might have specified a three-seat district in the northern part (covering the primary population centres of Nanaimo through Campbell River, and extending out to include the more remote communities of Tofino and Port Hardy) and a four-seat district in the southern part (covering the Greater Victoria region up to Duncan and including the Gulf Islands).
In the most recent federal election (2015), the NDP won about 33 percent of the vote, while the other three major parties (LP, CP and GP) each won 21-24 percent of the vote. But as our FPTP system would have it, the NDP took six of the seven seats, with the Green Party taking the seventh, as shown in Figure 3. The CPC and the Liberal Parties were shut out although they had garnered an appreciable share of the vote. Our simulations suggest that, under STV, had voters voted as they did in the 2015 election, the northern STV region would have elected MPs from three different parties – the NDP, the LPC and the CPC, while the southern STV region, with four seats, would have elected an MP from each of the four major contending parties (see Figure 3, which compares results under FPTP and STV).
This outcome would thus have much more fairly represented the actual votes cast. Looking at voter first choice preferences by party, STV would have allowed 94% of voters to select an MP in line with their first choice of party, vs. 40% under FPTP in 2015. This example illustrates how STV leads to a relatively high level of proportionality even with relatively small multi-member PR districts of only three and four seats.
This can be compared to MMP, where PR regions of eight are usually considered necessary to yield an acceptable level of proportionality. This is because MMP uses a winner-take-all system to elect local MPs (either FPTP or AV) and it takes a certain size of PR-region to correct the disproportionalities that result.
Central Toronto has 12 federal seats in Parliament and is one of the most heavily urbanized regions in the country. If we had conducted the 2015 election under STV, the EBC might have considered grouping all 12 ridings into a single electoral district, but this would be a larger district than Canada has ever used. Alternatively, they might have considered various combinations: two districts of six seats, one of eight and one of four, or perhaps three four-seat districts. Using three four-seat local districts, for example, one district might cover south-central Toronto, one the western portion of the city, and the third the eastern portion.
In the 2015 election, the Liberal Party won 50 percent of the vote there, while the NDP won 25 percent and the Conservatives won 20 percent–yet the Liberal Party took all 12 seats. If we had used STV with three STV districts in the 2015 election, our simulations show that the LPC would have won two seats in each of the three STV districts and the NDP and CPC one each.
In Central Toronto as a whole, the LPC would have won six seats, and the NDP and the CPC three each (Figure 4). Again, this result would fairly reflect the way Toronto voters voted, with 96 percent of constituents electing the party of their choice, compared to 50 percent under FPTP.
Canada as a Whole
We also performed a simulation of what might have happened in Canada as a whole if we had used STV nationwide in the 2015 federal election. For this simulation, we used slightly larger districts overall – the average number of MPs per district was 4.3, with up to seven or eight MPs in some urban areas (including metro Toronto) and as few as two or three in rural parts of the country (including a single MP in Labrador).
As shown in Figure 5 below, the Liberal Party won just under 40 percent of the vote in that election, the Conservatives 32 percent, the NDP 20 percent, the Bloc five percent and the Greens three percent. The LPC took 54 percent of the seats, the CPC 29 percent, the NDP 13 percent, the Bloc three percent, and the Green Party a single seat.
Under STV, we estimate that the LPC would have won 43 percent of the seats, the CPC 33 percent, the NDP 19 percent, and the Bloc four percent. The Green Party would have won two seats. Although this would not have been a perfectly proportional result, 91 percent of voters would have helped elect an MP from the party of their first choice, ensuring that the composition of Parliament would more closely reflect how Canadians voted.
More importantly, all major parties would have won representation in virtually all regions in which they ran candidates – the CPC and NDP would have won seats in Atlantic Canada and Toronto, and the LPC would have won more seats in the Prairies. In this way, STV would have virtually eliminated the Balkanization of the country that our current voting system promotes.
Under this basic form of STV, the Green Party would elect only two MPs, despite having enough supporters nationwide to justify 11 or 12 MPs. Green Party supporters are under-represented because when district sizes average about four MPs, a candidate needs to win about 20 percent of the vote to get elected in an average-sized region. Of course, this would be mitigated by the increased vote share that the Green Party could expect, once the need for strategic voting was eliminated under STV.
However, STV would certainly make it difficult for Green voters to elect representatives in areas of the country with smaller districts, and where support for the party is less than 10 percent. However, there are strategies that allow smaller parties to break through in selected STV ridings, for example by running only one star candidate, to avoid early elimination and capitalize on second order preferences
This limitation would also apply to larger parties with weak support in particular regions of Canada, particularly in rural regions where the number of seats per district would be lower. While the imperfect proportionality of the model is most clearly seen in the results for the Green Party in our simulation, one of the larger parties would likely also fail to win a seat in some of the smaller districts.
That said, there are relatively simple enhancements that could be made to STV (such as the small top-up layer described below) to enable supporters of smaller parties to combine their support across larger regions and elect an MP they prefer, thereby producing an overall more proportional result. Such a design feature could be well worth considering as a way to increase the overall equity of the STV model for all voters in all parts of the country.
The Appeal of STV as an Electoral Reform Option for Canada
As a proportional system, STV offers all of the advantages of PR as discussed in the main part of this submission. We will not repeat those advantages here. However, STV may appeal to committee members for reasons that are specific to this model, presented here in terms of the five ERRE principles, as we did in the main text of this submission.
The five ERRE principles and STV
Effectiveness and Legitimacy
As we have seen, with some caveats, STV leads to a high level of vote-effectiveness, on the order of 90 percent, even with relatively small STV ridings. It also offers voters more choice than virtually any other system. This includes:
- being able to rank one’s ranked preferences,
- being able to vote for individual candidates rather than party lists, and
- the freedom to vote across party lines.
Independent candidates are treated on an equal footing to party-affiliated candidates under STV. Although proportionality is usually expressed in terms of the vote share of each party vs. its share of seats, STV’s candidate-centred orientation allows us to express proportionality in terms of how many voters have elected specific candidates they support. Since each candidate would be elected with roughly equivalent levels of support, STV can be considered to maximize the legitimacy of each elected MP.
Voter engagement is likely to be high under STV because of the great amount of choice that it gives to voters.
As we have seen, STV also does a good job of eliminating regional strangleholds by one party under winner-take-all models. It does this successfully even with relatively small STV regions.
Accessibility and Inclusiveness
STV is relatively easy to understand because it is not a mixed system. All MPs are elected the same way, using ranked ballots.
Canada has a long history of using STV successfully, and no significant concerns have historically been raised about performing the count manually (though some counting rule variants do involve more handling of ballots than others, which can lead to longer counting times before final election results can be posted).
Even if more involved counting rules are used, it is usually possible to identify the likely winners with high certainty in the rounds of counting performed on the eve of the election, and thus the overall result of the election. The final determination of a small percentage of seats would normally be made after the eve of the election when all the ballot transfers have been completed (and if any postal votes are to be incorporated).
If it is important to have final election results produced more quickly, we might use optically scanned ballots (as are used in Vancouver’s civic election). This produces a paper trail for verification and the ballots can be independently counted by hand for the first round in the time that it currently takes to count our single-member election results. The optically-read ballots could be processed virtually instantly and the final results announced rapidly. Ballot details could be released so that independent observers could check the results of the counting program.
Local representation under STV is provided as part of multi-member districts. Although this formula may not seem just as “local” as single-member winner-take-all districts, voters would actually have more choice than ever about whom to consult, should they wish to speak to one of their MPs, and may prefer to speak with an MP for whom they actually voted. The ratio of MPs to voters would remain the same under STV.
In terms of accountability, each MP will be competing for votes with the other MPs to help the reelection chances of their party. Indeed, they may be competing also with possible contenders from their own party. MPs are fully accountable to voters in this system.
Admittedly, it is challenging to deal with ridings that have been enlarged in order to make them multi-member, particularly in rural areas where geographical distance is already an issue. This is the primary reason why the size of ridings in terms of the number of MPs they contain would normally be lower in rural areas, the lower limit being two-member ridings and even a handful of single-member ridings.
How much of a challenge this might be in practice is debatable, however. For instance, each major party is liable to run as many candidates as there are MP positions to be filled in each riding. This would allow these candidates to focus their campaigns on different parts of the multi-member riding. Similarly, with several MPs servicing the multi-member constituency, some division of labour might well emerge in ways that make geographical sense.
Using STV in 2019 could significantly reduce the cost of redistricting compared to some other options, since multi-member districts could often be created by simply regrouping existing ridings, as assumed in the above examples.
STV compared to other models for small region PR
STV is not the only way of managing multi-member elections in small regions. Other options are list-PR (with closed, flexible or open lists) and Stéphane Dion’s P3 proposal. These options provide all of the advantages of PR but do not offer voters the same degree of choice as STV.
List-PR with closed lists are not necessarily “undemocratic” because nominations could be done democratically and there are advantages to closed lists, which allow parties to balance the ticket that they put forward in terms of gender and other dimensions of diversity. However, they do not allow voters to choose among the candidates of a party. Flexible or open list PR go a step further by allowing voters to vote directly for individual candidates.
List-PR systems allow each voter one vote, without recourse to preferential ballots. This could work against candidates who are well liked by a significant share of the population but who may not be the first choice of most voters supporting that party. What can happen is that most people may vote for a party’s “star” candidate and have no opportunity to express their second preferences. The vote still counts for the party, but voter choice among candidates is not expressed as fully as it could be. Small parties may also be disadvantaged in list-PR systems with small regions, because the option of winning second-preference votes disappears.
Concerns have also been expressed about how open-list PR could affect party unity since candidates from the same party would be running against each other. STV helps deal with that problem because candidates of the same party usually ask voters to rank their colleagues of the same party next. This strategy maximizes the number of seats the party will win in each riding. With STV, candidates will work to secure second-choice preferences, including from those voters who favour their party but might rank another candidate first. This encourages cooperative and more supportive behaviour between candidates of the same party during an election campaign.
Stephane Dion’s P3 system calls for the use of a preferential ballot precisely for this reason. As Dion explains:
That means that candidates of the same party would compete for seats, which could undermine party cohesion. But with this new kind of competition, the candidates would still have to act as team members. They would have to show cohesion for their own party to be able to rally the votes needed to obtain seats. Those parties best able to combine cohesion and internal competition would have the best chances of winning — to the benefit of Canadians.
P3 operates in two steps. Step one uses voters’ preferences among parties to determine the allocation of seats. If the party of a voter’s first choice is eliminated at this stage, the voter’s second preference will kick in. The second step is to consider voters’ preferences among the party’s candidates to see which ones are elected. The difference with STV is that voters cannot cross party lines in expressing their preferences. Indeed, it is only voters for party X that determine who is chosen to fill party X’s spot or spots in the riding; voters whose first choice of party was eliminated do not get a personalized vote for candidates. Nor does Dion’s system make any allowance for surplus votes to be redistributed.
Dion’s system is “proportional, personal and preferential” as the name “P3” implies and is well worthy of consideration. However, it is more “party-centric” than STV. It does not count every voter’s second-preferences equally, does not allow preferences to be expressed across party lines, and does not readily allow for independents to compete on an even keel.
Of course, any system for electing candidates in relatively small multi-member ridings makes it difficult for third or fourth placed parties to win seats. This is true for list-PR, P3 or STV with regions of the same size. The limited size of multi-member ridings constrains the level of proportionality that can be achieved. This would affect the chances of small parties such as the Green Party everywhere, and of all parties some of the time, most notably in rural areas, where the multi-member ridings may have as few as two seats or even some single-member ridings.
While it is in principle possible to increase the size of multi-member districts to as much as 10 or 15 MPs to enable supporters of smaller parties to elect their preferred candidates, this is not a realistic option outside Canada’s largest cities, and would complicate things for voters, who would have to deal with very large ballot lists. Another option is to envisage a hybrid system along the lines of the Rural-Urban Proportional (RUP) system presented in Appendix 12.
Applied as an extension of STV, RUP would retain the use of STV in multi-member ridings while adding a certain number of top-up seats on a regional basis, as one does under MMP. We call this model STV+. Since STV is capable of delivering quite highly proportional results to begin with in most cases, only a few compensatory seats would be needed to fine-tune the system.
In Canada, the top-up formula could be applied regionally at an average rate of two top-up seats for every 13 regular STV MPs (or one top-up seat for every six regular STV seats in a few smaller areas). The actual size of top-up regions would vary according to geographic realities. On Vancouver island, it would be logical to have a top-up region of seven seats (6+1). In most other regions, it could be fifteen (13 + 2), including multi-member ridings of different sizes, any single-member ridings as appropriate and two top-ups, for a total of 15.
Our simulation exercise as shown in Figure 5 above suggests that had we used STV+ in the last election, the various parties would have ended up with a share of the seats approximating their share of the vote. The Liberals would be down 1% of the seats compared to regular STV, and the Green Party would now have won 2% of the seats instead of 0.6% (eight seats instead of two). Overall, the results would have more closely reflected voter intentions compared to STV alone. Please note that these numbers understate the improvement obtained by this innovation, since most improvements would take the form of increased proportionality at the regional level.
STV+ is flexible enough to accommodate a certain number of single-member ridings in particularly remote regions such as the northern parts of the western provinces, Ontario and Quebec, or Labrador. These single-member ridings would be included in top-up regions containing some multi-member ridings. Votes that may have been too few to elect a candidate at the riding level would count to elect a top-up candidate at the regional level, using the same proportionality principle as MMP.
STV+ could be designed to work well with no change to the number of MPs we currently have, but this would require redistricting to make room for the top-up MPs. Another option would be to preserve existing riding boundaries and add a number of new seats instead. We estimate that 50 new seats would be sufficient to play this top-up role.
In summary, we find that STV offers considerable appeal as a PR option for Canada. Although it has not been discussed as widely as MMP in recent years, experience to date in Canada and other Westminster democracies demonstrates STV’s technical feasibility and accessibility to voters. It is worth noting that there have been two referendums in Ireland about whether or not to keep STV, with the Yes side winning in both cases.
Fair Vote Canada suggests that STV and STV+ should be seriously considered as contenders by the Electoral Reform Committee, either as a stand-alone STV system, or using an STV+ formula combining small STV ridings and a certain number of top-up seats MMP style, as discussed in Appendix 12.
1 J. Patrick Boyer (1992). Direct Democracy in Canada: The History and Future of Referendums.
2 Justice Thomas Berger (2004). A City of Neighbourhoods: Report of the 2004 Vancouver Electoral Reform Commission. pp. 89-91